TW-I-01: The ROC Constitutional Architecture — The Five-Yuan System, the Seven Amendment Rounds, and the Politics of a Constitution That Can No Longer Be Amended (1947–2026)

Status: DRAFTWords: 11,660

1. Key Takeaways

Taiwan is governed under the world's only operating Sun Yat-sen constitution — a five-power design that grafts two indigenous Chinese institutions, the examination system and the censorate, onto the Montesquieuan trinity — and the document was written for a country of which Taiwan was then one province. The Constitution of the Republic of China, adopted by the Constituent National Assembly in Nanjing on 25 December 1946 and in force from 25 December 1947, implements Sun Yat-sen's wuquan xianfa (五權憲法, five-power constitution): an Executive Yuan, Legislative Yuan, and Judicial Yuan corresponding to the Western separation of powers, plus an Examination Yuan (考試院) institutionalising the imperial civil-service examination tradition and a Control Yuan (監察院) institutionalising the censorate (御史) tradition of supervising officialdom. Above the five governing powers (zhiquan, 治權) the design placed the political powers (zhengquan, 政權) of the people, exercised through a National Assembly (國民大會). Drafted with decisive input from the Democratic Socialist Party constitutionalist Carsun Chang (Zhang Junmai), who softened Sun's presidential design toward cabinet government, the constitution was boycotted by the Chinese Communist Party at adoption — and within two years its drafting state had lost the mainland. The strangeness of Taiwan's constitutional history begins here: a constitution written for 450 million people across thirty-five provinces has, since 1949, governed in practice an island its drafters barely discussed.
For its first forty-four years the constitution was suspended in practice almost as soon as it took effect: the Temporary Provisions of 1948 and the martial-law decree of 1949 built a parallel emergency state on top of the constitutional text. The Temporary Provisions Effective During the Period of National Mobilisation for Suppression of the Communist Rebellion (動員戡亂時期臨時條款), adopted 10 May 1948 — less than five months after the constitution took effect — gave the president emergency powers exempt from ordinary constitutional procedure, and a 1960 amendment exempted Chiang Kai-shek from the two-term presidential limit, which he exceeded until his death in 1975. The Taiwan martial-law decree of 19 May 1949 (lifted 15 July 1987; cross-ref TW-B-01) subordinated civil life to the Taiwan Garrison Command. The third pillar of the parallel state was the "ten-thousand-year parliament" (wannian guohui, 萬年國會): National Assembly delegates and Legislative Yuan members elected on the mainland in 1947–48 held their seats without re-election for over four decades, on the theory — ratified by the Council of Grand Justices in Interpretation No. 31 (1954) — that new national elections were impossible while the mainland constituencies were "occupied." The constitution thus survived as form while a Leninist party-state governed in substance; the 228 Incident (TW-J-01) and the White Terror unfolded under, not despite, this architecture.
Democratisation proceeded through the constitution rather than against it: seven rounds of Additional Articles between 1991 and 2005 rebuilt the ROC constitutional machine into a Taiwan-scaled semi-presidential democracy without ever replacing the 1947 text. Lee Teng-hui's method — forced partly by the Council of Grand Justices' Interpretation No. 261 (21 June 1990), which ordered the senior mainland-elected parliamentarians to retire by 31 December 1991 — was amendment by "Additional Articles" (增修條文) that suspend rather than delete the original provisions, preserving the formal claim that the constitution remains the constitution of all China while confining its operation to the "free area." The seven rounds delivered, in sequence: termination of the Temporary Provisions and full re-election of the parliament (1991); restructuring of the Control Yuan out of the parliamentary chamber system (1992); direct presidential election (1994, first held 1996); the semi-presidential settlement, with the premier appointed by the president without Legislative Yuan confirmation, and the freezing of the Taiwan Provincial Government (1997); a void self-extension of the National Assembly's term (1999, struck down by Interpretation No. 499 in 2000); the Assembly's reduction to a mission-convened rump (2000); and its abolition, the halving of the Legislative Yuan, and the referendum-ratification amendment rule (2005). No new constitution was ever promulgated — a deliberate choice with cross-Strait stakes — and the result is a constitutional order that is simultaneously revolutionary in content and conservative in form.
The 1997 settlement created a semi-presidential system of the president-dominant type, and thirty years of practice have confirmed the design's central asymmetry: the president appoints the premier freely, but no president has ever appointed an opposition premier, and the legislature has never dared pass a no-confidence vote. The Fourth Additional Articles round (July 1997) removed the Legislative Yuan's confirmation power over the premier, gave the LY a no-confidence weapon, and gave the president a counter-weapon of dissolution following a successful no-confidence vote. In the Shugart-Carey typology this is a president-parliamentary configuration; in practice it has produced what Taiwanese scholars call a president-dominant system in which the premier serves as the president's chief of staff and political fuse. Under divided government — Chen Shui-bian's full eight years (2000–2008) and the Lai era since February 2024 (cross-ref TW-A-03, TW-D-02) — the system generates sustained executive-legislative confrontation with no cohabitation valve: opposition legislators decline to move no-confidence because dissolution would put their own seats at risk, and presidents decline to appoint opposition premiers because nothing compels them to. The French-style alternation of cohabitation has never once been tested in Taiwan, and the 2024–2026 confrontation is the price of that untested seam.
The 2005 amendment round locked the constitution shut: the requirement that any amendment win referendum approval from an absolute majority of all eligible voters — not of votes cast — has made formal constitutional change practically impossible, as the 2022 voting-age referendum demonstrated. The June 2005 National Assembly, convened to ratify the 2004 Legislative Yuan proposal and then abolish itself, transferred amendment ratification to the electorate under a double lock: proposal by one-quarter of LY members and passage by three-quarters of a three-quarters quorum, then ratification by referendum in which affirmative votes must exceed 50 per cent of the total electorate. The only amendment ever to reach that referendum stage — the 26 November 2022 proposal to lower the voting age from 20 to 18, passed by the LY with near-unanimous cross-party support — won approximately 5.65 million yes votes against 5.02 million no, but fell roughly four million votes short of the required threshold of about 9.62 million [TBD-VERIFY: precise vote totals and threshold figure per Central Election Commission returns]. A constitution that cannot pass its own least controversial amendment cannot realistically pass any other; every structural reform debate since 2005 — abolishing the minor Yuans, lowering the voting age, redesigning the electoral system, clarifying the semi-presidential seams — runs into the same iron cage.
The Control Yuan and Examination Yuan — the censorate and examination inheritances — survive in shrunken, contested form, perpetually slated for abolition that the amendment rule guarantees will never come. The Control Yuan, converted by the 1992 round from an indirectly elected parliamentary chamber into a 29-member quasi-judicial body nominated by the president with legislative consent, retains impeachment, censure, corrective-measure, and audit powers, and since 2020 houses the National Human Rights Commission. Its modern career is defined by appointment warfare: the pan-Blue legislative majority refused to process Chen Shui-bian's nominees from February 2005 until 2008, leaving the body entirely vacant for over three years, and each subsequent appointment round has been a partisan battlefield. The Examination Yuan, custodian of the civil-service examination and personnel-protection system, was cut from 19 members to a much smaller bench with shortened terms in the 2020 reform round [TBD-VERIFY: post-2020 membership figure, commonly cited as 7 plus president and vice-president, and the term reduction from six to four years]. Both major parties have at various points endorsed a three-power consolidation — the DPP programmatically since the 1990s, and the KMT opportunistically in 2024–2025 when the Control Yuan declined to serve its purposes — yet abolition requires the amendment procedure that the 2005 rules foreclose, leaving budget strangulation and nomination blockade as the available substitutes for constitutional reform.
The Judicial Yuan's Constitutional Court has been both the quiet engine of Taiwan's constitutional development and, since 2024, the principal casualty of divided government. The Council of Grand Justices' interpretations drove democratisation at key junctures — No. 261 (1990) retiring the eternal parliament, No. 499 (2000) voiding the National Assembly's self-extension, No. 748 (2017) mandating same-sex marriage — and the Constitutional Court Procedure Act, effective 4 January 2022, converted the interpretation system into a true constitutional court issuing judgments and hearing individual constitutional complaints. The new architecture was then stress-tested almost immediately: in October 2024 the Court's Judgment No. 9 of 2024 struck the core of the KMT-TPP legislative-reform package; the opposition majority responded by rejecting all seven of President Lai's justice nominees in December 2024 and simultaneously amending the Procedure Act to require ten sitting justices for a hearing and nine votes for an unconstitutionality ruling — thresholds the eight-member rump bench could not meet, paralysing the Court by statute. The Court's partial strike-down of those thresholds in Constitutional Judgment No. 2 of 2025 (28 March 2025) [TBD-VERIFY: exact case number and date, per the TW-D-06 corpus line] restored constrained functionality, but the confirmation blockade has persisted through mid-2026, leaving constitutional review running at reduced capacity precisely when inter-branch disputes most require it.
The 2024–2026 divided-government cycle is the most severe constitutional stress test since direct presidential elections began, and it has exposed every unfinished seam of the 1997 and 2005 settlements at once. The sequence — the May 2024 legislative-reform package with its contempt-of-legislature powers and the Qingniao (Bluebird) street response; Judgment No. 9 of 2024; the December 2024 Constitutional Court Procedure Act amendment and justice-confirmation blockade; the January 2025 budget cuts and freezes of roughly NT$207.5 billion [TBD-VERIFY: aggregate figure, per TW-D-06]; the mass recall campaigns of 2025, whose July and August vote days failed in every district and left the opposition majority intact; and the FY2026 budget confrontation (all treated in TW-D-02, TW-D-06) — is, institutionally read, a single demonstration that Taiwan's constitution contains no settled answer to divided government. The no-confidence/dissolution mechanism sits unused; the budget rules let the legislature cut but not add, inviting strangulation rather than negotiation; the Constitutional Court's composition can be held hostage by the very majority whose acts it reviews; and the recall and referendum instruments, liberalised in 2016–2017, have become continuations of electoral warfare by other means. That the system has bent without breaking — courts still sitting, budgets still passing, recalls resolved by voting — is the optimists' evidence; that every escalation rung has now been used short of the last is the pessimists'.
The deepest constraint on constitutional reform is not procedural but existential: the ROC constitution is simultaneously Taiwan's democratic vessel and the legal anchor of "one China," so changing it formally is read in Beijing — and in Washington — as touching the sovereignty question. The Additional Articles' preamble frames every amendment as serving "the requisites of the nation prior to national unification"; the territory clause was never amended to match the governed reality; the Act Governing Relations Between the People of the Taiwan Area and the Mainland Area (1992) statutorily encodes the one-country-two-areas frame. This is why Chen Shui-bian's 2003–2006 push for a "new constitution" alarmed Washington as much as Beijing, why the PRC's 2005 Anti-Secession Law treats constitutional change as a potential casus belli trigger, and why the independence movement faces a standing dilemma: the constitution it would replace is also the only constitutional order under which Taiwan's democracy exists, and the PRC's own legal position paradoxically depends on that constitution's "one China" residue remaining in place. The amendment iron cage of 2005 is therefore double-locked — procedurally by the referendum threshold, strategically by the cross-Strait status quo (cross-ref TW-F-02, TW-K-01, TW-K-03) — and Taiwanese constitutional politics has adapted by changing everything except the text: through interpretation, statute, practice, and electoral mandate.
Read comparatively, Taiwan's constitutional order is an extreme case on two dimensions simultaneously: it is among the world's most flexible constitutions in practice (seven major revisions in fourteen years) and among the most rigid in current law (no amendment possible in twenty years and counting) — and both readings of that rigidity are at full strength. The stability reading holds that the post-2005 freeze has insulated the fundamentals — direct presidential election, the human-rights jurisprudence, the cross-Strait status quo — from the volatility of Taiwan's intense partisan competition, and that the 2024–2026 crisis was contained precisely because neither side could change the rules. The ossification reading holds that a constitution that cannot incorporate an 18-year-old voting age, cannot resolve its semi-presidential ambiguities, cannot rationalise its five-branch redundancy, and cannot align its formal territory with its actual democracy is accumulating legitimacy debt that will eventually be called in. The honest position, as of mid-2026, is that both are true: Taiwan operates a 1947 mainland constitution, amended through 2005, governing a 2026 island democracy — an arrangement that is at once the region's most improbable constitutional success and its most obviously unfinished one.

2. The 1947 Constitution and Its Strange Career

2.1 Sun Yat-sen's Five-Power Design

The intellectual architecture of the ROC constitution is Sun Yat-sen's, set out in his Three Principles of the People (三民主義, Sanmin Zhuyi) lectures and his Fundamentals of National Reconstruction (建國大綱, 1924). Sun's constitutional theory made two moves that distinguish it from every other twentieth-century constitutional tradition. First, he separated the people's political powers (zhengquan, 政權) — election, recall, initiative, and referendum — from the government's governing powers (zhiquan, 治權), arguing that a strong, expert government could be reconciled with popular sovereignty if the people held the four political powers over it. Second, he expanded Montesquieu's three governing powers to five, adding two institutions he regarded as China's distinctive contributions to the science of government: an independent examination power, descending from the imperial civil-service examination system (科舉, keju) that had recruited the Chinese bureaucracy for thirteen centuries, and an independent supervisory power, descending from the censorate (御史臺 / 都察院) that had audited and impeached officials on the emperor's behalf since the Qin. Sun's stated reasoning was that lodging examination inside the executive invited patronage, and lodging impeachment inside the legislature invited the congressional abuse he observed in the United States — legislatures using oversight as partisan warfare. Whether the five-power design solved those problems or merely relocated them is the central question of this document's Sections 5 and 6.

Sun also prescribed a three-stage sequence to constitutional government: military unification (軍政), political tutelage (訓政) under Kuomintang guidance, and constitutional rule (憲政). The tutelage stage, formalised in the 1931 Provisional Constitution of the Political Tutelage Period, made the KMT the legal trustee of the people's political powers — a doctrine whose afterlife on Taiwan would be long.

2.2 Nanjing, 1946: Drafting a Constitution for All China

The constitution finally adopted on 25 December 1946 was not the KMT's preferred document. The party's 1936 "May Fifth Draft" (五五憲草) had envisaged a powerful presidency atop the five Yuans. The 1946 Political Consultative Conference — the short-lived postwar negotiation among the KMT, the Chinese Communist Party, the Democratic League, and minor parties — extracted commitments to a more parliamentary design, and the constitutional text that emerged bore the decisive imprint of Carsun Chang (Zhang Junmai, 張君勱), the German-trained Democratic Socialist Party constitutionalist who served as principal drafter. Chang's compromise grafted cabinet-government features onto Sun's framework: the Executive Yuan was made responsible to the Legislative Yuan (Article 57 of the original text), the premier's appointment required legislative consent (Article 55), and the presidency was framed as a largely mediating headship of state. Sun's National Assembly survived as the organ of the people's political powers — electing and recalling the president and vice-president and amending the constitution — but as an occasional body rather than a standing super-parliament.

The Constituent National Assembly that adopted the text did so without the CCP and the Democratic League, both of which boycotted after the breakdown of the Marshall mediation; the constitution was thus born already an artefact of the civil war it could not survive. It entered into force on 25 December 1947, with the first National Assembly, Legislative Yuan, and (indirectly elected) Control Yuan chosen in the 1947–48 elections held across a mainland already fracturing. Within a year of the first president's inauguration the constitutional state was losing the war; by December 1949 its government sat in Taipei, governing Taiwan and a scattering of offshore islands — including Kinmen and Matsu, the only fragments of its claimed Fujian province it retained — under a constitution designed for thirty-five provinces.

It should be recorded, because it shapes everything after, what Taiwan's place in this founding was: the island had been retroceded from Japan only in October 1945, was governed not by ordinary provincial administration but by Chen Yi's Administrative Executive Office, and in the very months the constitution was being adopted and promulgated was experiencing the 228 Incident and its suppression (TW-J-01). Taiwanese participation in the constitution's drafting was marginal — a handful of delegates in an assembly of thousands. The constitution that would eventually become the vessel of Taiwanese democracy arrived on the island, in the candid summary of much post-democratisation scholarship, as the legal instrument of an externally imposed regime.

2.3 The Temporary Provisions and the Parallel State (1948–1991)

The constitutional era proper lasted, on the most generous count, 136 days. On 10 May 1948 — before the first president had even taken office — the National Assembly adopted the Temporary Provisions Effective During the Period of National Mobilisation for Suppression of the Communist Rebellion (動員戡亂時期臨時條款), using the constitution's own amendment procedure to exempt the president from the constitutional constraints on emergency powers. The Temporary Provisions were repeatedly expanded after the move to Taiwan: a March 1960 revision suspended the two-term presidential limit of Article 47, allowing Chiang Kai-shek's election to a third term that year and ultimately to a fifth (he died in office in 1975, succeeded constitutionally by Vice-President Yen Chia-kan and politically by his son Chiang Ching-kuo, elected president in 1978); 1966–1972 revisions authorised the president to create the National Security Council apparatus by decree and to add "supplementary" parliamentarians from the Taiwan area without disturbing the mainland-elected core.

The second layer of the parallel state was martial law, declared for Taiwan on 19 May 1949 and maintained for thirty-eight years (lifted 15 July 1987; cross-ref TW-B-01) — at the time the longest continuous martial law in modern history. Under it, the Taiwan Garrison Command exercised censorship, ban powers over assembly and new political parties (the dangjin, 黨禁, and baojin, 報禁), and military trial of civilians for sedition under the 1949 Act for the Punishment of Rebellion and the 1950 Statute for the Detection and Suppression of Communist Spies — the statutory machinery of the White Terror.

The third layer was the "ten-thousand-year parliament" (萬年國會). The 1947–48 mainland-elected National Assembly delegates, Legislative Yuan members, and Control Yuan members could not be re-elected once their constituencies fell to the Communists; rather than concede that the parliament had lapsed, the Council of Grand Justices held in Interpretation No. 31 (1954) that the incumbents should continue to serve until re-election became possible — which, on the regime's own premises, meant until the mainland was recovered. The first-term parliamentarians therefore sat for over four decades, dying gradually in office, their dwindling ranks topped up from 1969 onward by limited "supplementary elections" (增額選舉) in the Taiwan area that gave the dangwai (黨外, "outside-the-party") opposition its only electoral arena. The constitutional symbolism was perfect and perfectly corrosive: the legislature's mainland composition was the daily, visible embodiment of the claim that the ROC was all China — and of the corollary that the Taiwanese governed could not choose their governors.

The composite result is what Taiwanese constitutional scholarship calls the "frozen constitution" inheritance. The 1947 text was never abolished — the regime's legitimacy claim required it — but it governed almost nothing important. When democratisation came, therefore, it could be framed not as revolution but as defrosting: returning to a constitution that had always formally been there. That framing was Lee Teng-hui's central strategic asset, and its costs — the preservation of the all-China form, the five-Yuan structure, and the mainlander-drafted text as the permanent baseline — are Taiwan's constitutional politics to this day.

3. The Seven Amendment Rounds (1991–2005)

3.1 The Forcing Events: Interpretation No. 261 and the National Affairs Conference

The amendment era had a judicial trigger and a political one. The judicial trigger was the Council of Grand Justices' Interpretation No. 261 of 21 June 1990, which — reversing the spirit of its own Interpretation No. 31 — ordered all remaining first-term (mainland-elected) parliamentarians to leave office by 31 December 1991, holding that periodic re-election was a constitutional requirement that the changed circumstances no longer excused. The political trigger was the March 1990 Wild Lily student movement (野百合學運), tens of thousands occupying what is now Liberty Square to demand the National Assembly's dissolution, direct elections, and a timetable for reform — pressure Lee Teng-hui, elected to a full presidential term by that same senescent Assembly weeks earlier, converted into the June–July 1990 National Affairs Conference (國是會議), the cross-party conclave (including the DPP, legalised only in 1986–89) that produced the reform roadmap.

Lee's method, settled at this juncture, deserves emphasis because it determined the constitutional order's shape: reform would proceed by Additional Articles (增修條文) adopted under the constitution's own Article 174 procedure, which would suspend specified provisions of the 1947 text "to meet the requisites of the nation prior to national unification" (因應國家統一前之需要) — the preamble formula that still heads the Additional Articles. The 1947 constitution would be neither replaced nor formally amended in its body; Taiwan would be carved out as the constitution's operative zone (the "free area," 自由地區) without renouncing the whole. The alternative path — a new constitution, openly Taiwanese — was the DPP's demand throughout (its 1991 draft "Constitution of the Republic of Taiwan" accompanied its independence-clause party charter) and was rejected precisely because of its cross-Strait meaning.

3.2 Rounds One to Three: Ending the Old Parliament and Winning Direct Election (1991–1994)

The First round (April 1991), passed by the outgoing first-term National Assembly as its final act, terminated the Period of National Mobilisation (President Lee declared it ended 1 May 1991, simultaneously abolishing the Temporary Provisions), provided the electoral basis for an all-Taiwan second National Assembly (elected December 1991) and second Legislative Yuan (elected December 1992 — the first fully re-elected parliament in the ROC's history), and grandfathered the cross-Strait relationship into law (authorising what became the 1992 Act Governing Relations Between the People of the Taiwan Area and the Mainland Area).

The Second round (May 1992) restructured two of the five powers: the Control Yuan ceased to be an indirectly elected parliamentary chamber and became a quasi-judicial body of members nominated by the president with National Assembly consent (Section 5.1 below), and the provincial governor and the mayors of Taipei and Kaohsiung were made directly electable. It also fixed the presidential term at four years and committed, ambiguously, to popular presidential election "by the entire population" — deferring the fight between direct election (Lee's faction and the DPP) and an American-style or assembly-mediated formula (the KMT's "non-mainstream" mainlander faction) to the next round.

The Third round (July 1994) resolved that fight for direct election: the president and vice-president would be elected "by the entire populace of the free area" on a single ticket, by relative majority, from 1996. The first direct presidential election of 23 March 1996, conducted under PRC missile bracketing of the island's ports, returned Lee with 54 per cent and completed the transition's headline arc (the corpus's own scope marker; see taiwan/CLAUDE.md). The 1994 round also pruned the premier's countersignature power over presidential appointments — the first cut in a progressive presidentialisation of the Chang Junmai cabinet design.

3.3 Round Four (1997): The Semi-Presidential Settlement and the Freezing of the Province

The Fourth round (July 1997) is the constitutional settlement Taiwan still lives under, and it was a genuine KMT-DPP bargain — negotiated between Lee and DPP chairman Hsu Hsin-liang's party at a second conclave, the December 1996 National Development Conference, over the objection of substantial factions in both parties. Its elements:

The premier (President of the Executive Yuan) is appointed by the president without Legislative Yuan confirmation, deleting the original Article 55 consent requirement. This was the round's pivot: Lee had been forced into bruising confirmation fights (the LY's 1996 resistance to reappointing Premier Lien Chan while he remained vice-president-elect), and the DPP — calculating that it could win the presidency before it could win the legislature — preferred a strong presidency to a strong parliament.
The Legislative Yuan received a no-confidence vote against the premier (by simple majority on the signature of one-third of members), and the president received dissolution power over the LY, exercisable only after a successful no-confidence vote. The pairing was modelled loosely on the French Fifth Republic, but with the dissolution trigger conditional rather than free — a difference with large consequences (Section 4).
The Executive Yuan's veto-like mechanism was recalibrated: the EY may, with presidential approval, return a bill to the LY, which may uphold it by an absolute majority of all members (lowered from the original two-thirds) — after which the premier must accept it, the original text's option of resignation being deleted.
The Taiwan Provincial Government was "frozen" (精省, jingsheng — "streamlining the province"): the elected governorship and provincial assembly, created only in 1992 and filled in the December 1994 elections, were abolished, the province reduced to a nominal commission under the Executive Yuan. The official rationale was the near-total overlap between a province covering over 95 per cent of the governed territory and the central government above it. The political subtext was personal and consequential: the sitting elected governor, James Soong (宋楚瑜), had won 4.7 million votes in 1994 — a mandate rivalling the presidency — and was emerging as a rival to Lee's chosen succession. The freezing destroyed Soong's platform; his breakaway independent presidential run in 2000 split the pan-Blue vote (Soong 36.8 per cent, Lien Chan 23.1) and delivered the presidency to Chen Shui-bian with 39.3 per cent — after which Soong founded the People First Party. The provincial government lingered as a vestige until its remaining budget and functions were finally zeroed out under the Tsai administration in 2018–2019. No single episode better illustrates how Taiwan's constitutional engineering and its elite power struggles were the same events described twice.

3.4 Rounds Five and Six: The National Assembly Destroys Itself (1999–2000)

The Fifth round (September 1999) is remembered as the system's self-dealing nadir: the third National Assembly voted to extend its own term (and the Legislative Yuan's) by over two years, aligning future elections but incidentally postponing its members' own accountability. The Council of Grand Justices responded with Interpretation No. 499 (24 March 2000), voiding the amendment in its entirety — holding that amendments adopted by secret ballot in violation of procedure, and amendments violating the constitution's foundational principles of democratic accountability, were beyond even the amending power. Interpretation No. 499 is Taiwan's basic-structure moment, establishing that the amendment power has substantive limits — a doctrine with obvious, never-tested implications for any future amendment touching sovereignty.

Disgraced, the Assembly in the Sixth round (April 2000) reduced itself to a "mission-oriented" (任務型) body: no longer standing, to be elected by proportional representation only when needed to ratify LY-proposed constitutional amendments, territorial alterations, or presidential impeachments. Its other powers migrated to the Legislative Yuan — including consent over Judicial, Examination, and Control Yuan appointments, which is how the LY acquired the confirmation weapons it would use in 2005–2008 and again in 2024–2026.

3.5 Round Seven (2005): The Halving, the New Electoral System, and the Iron Cage

The Seventh round (June 2005) was ratified by the last National Assembly ever convened — 300 members elected in May 2005 by party-list PR solely to approve the package the Legislative Yuan had passed in August 2004 — which then abolished itself. The package had three structural elements, each with effects its sponsors did not fully anticipate:

The Legislative Yuan was halved from 225 to 113 seats (effective with the 2008 election), with terms lengthened from three to four years. "Halving the legislature" (國會減半) was a populist cause embraced across party lines in the post-2004 anti-politician mood; its critics — heavily represented in the academy — warned that 113 seats for 23 million people would produce one of the democratic world's highest constituent-to-representative ratios and thin committee expertise. The 2024–2026 confrontations, in which a three-seat swing controls the chamber, are routinely cited in this literature's vindication file.
The electoral system became mixed-member majoritarian: 73 single-member districts decided by plurality, 6 indigenous seats, and 34 at-large seats by closed-list PR (with a 5 per cent threshold and a half-female list requirement), votes cast on two separate ballots without compensation. The system structurally advantages the two big parties in districts while giving small parties a narrow PR window — the configuration that produced the TPP's eight-seat kingmaker bloc of 2024 (TW-D-02).
Constitutional amendment was transferred to referendum ratification: amendments require proposal by one-quarter of LY members, passage by three-quarters of members present at a sitting with a three-quarters quorum, publication for six months, and then ratification at referendum by affirmative votes exceeding 50 per cent of the total number of eligible voters — not of votes cast. The same round moved presidential impeachment adjudication to the Constitutional Court and entrenched the abolition of the Assembly.

The third element is the iron cage. A threshold keyed to the whole electorate means that abstention counts as a no vote; with turnout in non-presidential-year referendums historically far below 75 per cent, the arithmetic demands a degree of cross-party mobilisation that Taiwan's polarised system cannot supply even for consensus items. The proof arrived on 26 November 2022, when the only amendment ever to clear the LY stage under the new rules — lowering the voting age from 20 (a 1947 figure unaltered since) to 18, passed in March 2022 with near-unanimous cross-party support — won a majority of votes cast (approximately 5.65 million yes to 5.02 million no) but reached only about 29 per cent of the electorate against the required threshold of roughly 9.62 million [TBD-VERIFY: precise CEC figures]. Taiwan thus retains the highest voting age among the world's established democracies because of an amendment rule, not a policy disagreement. President Tsai's 2020–2022 constitutional-reform committee process and President Lai's occasional gestures toward institutional amendment have both foundered on the same arithmetic, and every serious reform debate since — Yuan consolidation, semi-presidential clarification, even technical electoral fixes — begins and ends with it.

4. The Semi-Presidential Reality

4.1 The Triangle on Paper

The post-1997 system is conventionally classified as semi-presidential: a popularly elected fixed-term president coexists with a premier and cabinet responsible to the legislature — Duverger's defining couple, refined in Robert Elgie's minimal definition under which Taiwan is an unambiguous member of the family. Within the family, the Shugart-Carey distinction matters more: in premier-presidential systems (France, post-2015 variants) the cabinet is responsible to the assembly alone, while in president-parliamentary systems the cabinet answers to both president and assembly. Taiwan after 1997 is president-parliamentary: the president appoints the premier freely, removes him in practice at will (the constitutional text is silent on removal, but no premier has ever survived a president's wish that he go), and the LY may also bring him down by no-confidence. The comparative literature — Shugart and Carey themselves, and the Taiwan-focused work of Yu-Shan Wu, Jung-Hsiang Tsai, and Shelley Rigger — has long flagged president-parliamentarism as the family's accident-prone branch, because the premier serves two masters who may be enemies.

Three further design facts complete the triangle. First, the president's dissolution power is conditional: the LY can be dissolved only within ten days after it passes a no-confidence vote — so a legislature that simply refuses to use the no-confidence weapon can never be dissolved, however obstructive it is. Second, the LY's budget power is asymmetric: under Article 70 of the original text it may cut and freeze appropriations but may not increase expenditure, which channels divided-government conflict into strangulation rather than logrolling. Third, the president's own constitutional portfolio (national security, defence, foreign affairs, cross-Strait policy, via the National Security Council inherited from the Temporary Provisions era and constitutionalised in 1991) overlaps undemarcatedly with the Executive Yuan's position as "the highest administrative organ of the State" (Article 53) — a seam papered over whenever president and premier are co-partisans and exposed whenever the question "who decides?" is asked adversarially.

4.2 The Chen Years: Divided Government Mark I (2000–2008)

Chen Shui-bian's two terms (TW-A-03) were the system's first full divided-government experiment: the DPP never held more than a plurality in the LY, and the pan-Blue KMT-PFP bloc held a working majority throughout. The institutional record set the patterns that recur in 2024–2026:

The revolving-door premiership. Chen ran through six premierships in eight years — Tang Fei (a KMT figure appointed in a gesture of "government for all the people," gone within five months), Chang Chun-hsiung, Yu Shyi-kun, Frank Hsieh, Su Tseng-chang, and Chang Chun-hsiung again — with the premier functioning as the presidency's political fuse, replaced after each electoral setback or stalemate. The average tenure of premiers since 1997 is well under two years, against four-year presidential terms; the office's subordination is the system's clearest revealed fact.
The Fourth Nuclear Power Plant crisis (October 2000 – February 2001), the constitutional order's first inter-branch collision: Premier Chang announced suspension of the LY-budgeted plant, the pan-Blue majority threatened recall and no-confidence, and the Council of Grand Justices' Interpretation No. 520 (15 January 2001) held the unilateral suspension procedurally defective, requiring the EY to report to and seek a resolution of the LY. The episode established the Constitutional Court as the divided-government referee — and previewed the budget-and-referral choreography of 2024–2026.
The non-use of the no-confidence weapon. The pan-Blue majority, despite repeated threats (including motions tabled and defeated in this period), never brought Chen's premiers down — the standard explanation being that legislators elected under SNTV (and after 2008 in single-member districts) feared snap-election risk more than they valued an opposition premiership, particularly since dissolution would follow any successful motion. Only three no-confidence motions have ever reached a vote (1999 against Vincent Siew; September 2012 against Sean Chen; October 2013 against Jiang Yi-huah), and all failed [TBD-VERIFY: the full count of tabled-and-voted motions].
The Control Yuan vacancy (2005–2008), treated in Section 5.1 — the first demonstration that the LY's consent powers over the other Yuans could be used to switch an entire constitutional branch off.

The cohabitation question crystallised here and has never been answered since. French practice would have dictated that Chen, facing a hostile majority, appoint a pan-Blue premier and retreat to the presidency's reserved domains. Chen's Tang Fei appointment gestured at this and collapsed within months — Tang was a KMT individual, not a KMT government, and his party never accepted ownership. Thereafter no president has tried: the 1997 rules do not require it, the president's electoral mandate argues against it, and — the deepest reason — Taiwan's party division is not a normal policy cleavage but a sovereignty-identity cleavage, making the delegation of the government to the other camp feel, to each side's base, like surrendering the country rather than alternating the administration. Cohabitation remains the system's most important untested provision.

4.3 The Ma and Tsai Years: Unified Government as the System's Default Mask (2008–2024)

For sixteen consecutive years the triangle's seams were hidden by unified government: Ma Ying-jeou (2008–2016, TW-C-01) and Tsai Ing-wen (2016–2024) each governed with same-party LY majorities for their full tenures. The period demonstrated the system's other face — presidents commanding effectively majoritarian power, with the premier as implementation manager (Ma's Liu Chao-shiuan, Wu Den-yih, Sean Chen, Jiang Yi-huah, Mao Chi-kuo, Chang San-cheng; Tsai's Lin Chuan, William Lai, Su Tseng-chang, Chen Chien-jen) and the LY's institutional weakness visible in the opposite register: majority-party legislators processing executive agendas under party discipline enforced from the presidential office, with the speakership (Wang Jin-pyng under Ma — whose 2013 attempted purge by Ma triggered the "September strife" and previewed intra-camp constitutional warfare; You Si-kun under Tsai) as the only consistent counterweight. The 2014 Sunflower occupation of the LY chamber (TW-B-05, TW-K-01's era context) was, in institutional terms, an extra-constitutional check substituting for the intra-constitutional ones that unified government had idled — and its legacy, the demand for stronger legislative oversight of executive (especially cross-Strait) dealmaking, is among the genealogies the KMT-TPP bloc would invoke for the 2024 reform package.

4.4 The Lai Era: Divided Government Mark II (2024–)

The January 2024 elections returned the triangle to its adversarial configuration with a new geometry: Lai Ching-te won the presidency with 40.05 per cent while the DPP fell to 51 of 113 seats, against the KMT's 52 and the TPP's pivotal 8 (TW-D-01, TW-D-02). Lai appointed Cho Jung-tai premier and made no gesture toward coalition or cohabitation; the KMT-TPP bloc, for its part, has conspicuously declined to move no-confidence — preserving its majority from dissolution risk — while prosecuting the confrontation through the instruments Sections 5.3 and 6 describe: the May 2024 reform package, the justice-confirmation blockade, the Constitutional Court Procedure Act amendment, and the budget knife. The Lai-era record through mid-2026 (TW-D-06) is thus the Chen-era pattern at higher intensity and with a fully weaponised toolkit; what is genuinely new is the targeting of the referee institution itself (Section 5.3).

5. The Minor Yuans and the Reform Debates

5.1 The Control Yuan: The Censorate's Modern Career

The Control Yuan (監察院) is the censorate inheritance: in the original 1947 design an indirectly elected parliamentary chamber (chosen by provincial assemblies, treated by Interpretation No. 76 of 1957 as one of the three organs jointly equivalent to a Western parliament), converted by the 1992 Additional Articles into a quasi-judicial supervisory body of 29 members serving six-year terms, nominated by the president with the consent of the National Assembly — after 2000, of the Legislative Yuan. Its powers: impeachment (彈劾) of public functionaries including judges and military officers (impeachment of the president was moved to other tracks by the amendments), referred for sanction to the Judicial Yuan's Court of the Judiciary; censure (糾舉) of officials for prompt administrative discipline; corrective measures (糾正) directed at agencies; the receipt of citizens' petitions; the audit power, exercised through the subordinate National Audit Office; and, since the 1990s, administration of the public-functionary asset-declaration regime. Since 2020 it also houses the National Human Rights Commission — a Paris Principles body created by statute, with its first chair the Control Yuan president Chen Chu — the most significant expansion of the institution's mandate since 1992.

The institution's modern career is a contradiction. Functionally, it is Asia's oldest ombudsman-type body, with a caseload of investigations (Lafayette frigate procurement, 319 shooting aftermath, Hualon and SOGO scandals, transitional-justice-era reinvestigations of White Terror cases) that its defenders cite as exactly the politically insulated supervision Sun designed. Politically, it has been progressively delegitimised by its own appointment process. The decisive episode was February 2005 – July 2008: the pan-Blue LY majority refused even to calendar Chen Shui-bian's nominee slate, and the fourth-term Control Yuan simply never came into existence — the entire branch stood vacant for three and a half years, its pending cases frozen, until Ma Ying-jeou's 2008 nominations were confirmed. The Council of Grand Justices' Interpretation No. 632 (2007) held the LY's refusal to exercise its consent power unconstitutional in effect — and was simply not complied with, an early demonstration that interpretations without remedies depend on political good faith. Subsequent rounds have been only less extreme: Ma's 2014 slate saw eleven of twenty-nine nominees voted down by his own party's legislators; Tsai's 2020 slate, headed by Chen Chu, was confirmed amid physical brawls in the chamber; and the KMT-TPP majority's treatment of the institution since 2024 — slashing its operating budget in the January 2025 round to a degree the Control Yuan itself described as operationally disabling [TBD-VERIFY: the FY2025 cut percentages and the FY2026 disposition], alongside KMT proposals to abolish the body outright — has made the censorate a direct combatant in the divided-government war.

The abolition debate is as old as democratisation. The DPP's platform tradition, and a substantial scholarly school, has long favoured a three-power consolidation: fold impeachment and audit into the legislature or an independent ombudsman-audit commission, fold the Examination Yuan into the executive's personnel administration, and retire Sun's pentagon as redundant — noting that the Control Yuan's powers overlap the LY's investigative oversight, the prosecutors' corruption jurisdiction, and the Judicial Yuan's discipline court, and that bodies of presidential appointees disciplining the president's own administration invite either timidity or partisanship. The counter-case: the LY has never built serious investigative capacity (the 2024 attempt to acquire it is Section 6's subject); audit independence needs an institutional home; the NHRC needs a constitutional shelter; and the patronage utility of twenty-nine dignified sinecures is, unspokenly, valued by every president who holds the nomination pen. But the debate is academic in the precise sense: abolition requires constitutional amendment, the amendment requires the Section 3.5 iron cage to open, and so the actually existing politics of the minor Yuans is conducted through the substitutes — nomination blockade, budget strangulation, and rhetorical delegitimisation — that damage the institutions without ever resolving their status.

5.2 The Examination Yuan: The Examination State's Residue

The Examination Yuan (考試院) administers the inheritance of the keju: national civil-service examinations (through the Ministry of Examination), and the appointment, grading, pay, tenure, and pension architecture of the civil service (through the Ministry of Civil Service, with associated commissions for protection-and-training and pension administration). Its constitutional rationale was insulation of merit recruitment from spoils; its practical twentieth-century function on Taiwan included, less gloriously, the administration of provincial-quota examination rules that for decades structurally favoured mainlander candidates [TBD-VERIFY: the operation and abolition timeline of the mainland-province quota system in ROC examinations on Taiwan].

Its modern trajectory is shrinkage. The body's policymaking top — president, vice-president, and members, nominated by the president with LY consent — was cut by statute in 2020 from nineteen members to a configuration commonly reported as seven members plus the president and vice-president, with terms shortened from six to four years [TBD-VERIFY: the amended Organic Act figures and effective date, applied from the September 2020 term]. The 2017–2018 pension-reform fight (TW-K-02) ran partly through its turf. Its examinations remain the gate to one of Asia's more professionalised bureaucracies, but virtually no one defends its Yuan status on first principles any more: the standing reform consensus — endorsed in different keys by DPP platforms, KMT technocrats, and the public-administration academy — would relocate its functions to an independent civil-service commission under or beside the Executive Yuan. It survives for the same reason the Control Yuan does: the iron cage, plus inertia, plus the quiet usefulness of its appointments.

5.3 The Judicial Yuan and the Constitutional Court: From Interpreter to Combatant

The Judicial Yuan (司法院) presides over the ordinary courts, the administrative courts, and the discipline court, but its constitutional centre is the body the world knows as Taiwan's Constitutional Court. Its institutional evolution tracks the whole arc of this document: the Council of Grand Justices (大法官會議) of the authoritarian era legitimating the parallel state (Interpretation No. 31); the same Council as democratisation's instrument (No. 261 of 1990; No. 499 of 2000; No. 520 of 2001) and as the rights engine of the consolidated democracy (No. 748 of 24 May 2017, mandating same-sex marriage within two years — the basis of the 2019 legislation); and, since the Constitutional Court Procedure Act (憲法訴訟法) took effect on 4 January 2022, a true constitutional court. The 2022 architecture: fifteen justices serving staggered eight-year terms (a 2003-era reform ending renewability), nominated by the president with LY consent; named judgments (憲判字) with majority authorship and published votes replacing the old anonymous interpretations; and — the deepest change — an individual constitutional complaint allowing persons to challenge final court judgments, converting the body from an abstract referee among organs into a supreme rights tribunal for litigants.

The new court's stress test arrived within three years, and it is the institutional story of the Lai era. The sequence, treated chronologically in TW-D-02/TW-D-04/TW-D-06 and analytically here: the Court's Judgment No. 9 of 2024 (113年憲判字第9號, 25 October 2024) struck the core of the KMT-TPP legislative-reform package (Section 6.1). On 31 October 2024 the terms of seven justices — including the Judicial Yuan president and vice-president — expired; in December 2024 the LY majority rejected President Lai's entire seven-name replacement slate, leaving eight sitting justices, and in the same month amended the Constitutional Court Procedure Act to require ten participating justices for oral argument and judgment and nine affirmative votes for any holding of unconstitutionality — thresholds arithmetically unreachable by an eight-member bench. The amendment was, on its face, a quorum rule; in operation it was a statute switching off constitutional review, enacted by the majority whose acts were under review, while the same majority blocked the appointments that would have restored quorum — the move scholars across the spectrum compared to the Polish and Hungarian court-capture sequences, while the bloc's defenders framed it as a legitimate check on a DPP-appointed bench [TBD-VERIFY: the amendment's precise thresholds and vote date]. In Constitutional Judgment No. 2 of 2025 (114年憲判字第2號, 28 March 2025) [TBD-VERIFY: exact case number and date, per the TW-D-06 corpus line], the rump bench partially struck the amendment, restoring its own operative functionality at eight members; the LY has refused every subsequent renomination through the May–June 2026 horizon, and the Court has operated at reduced capacity and tempo since (TW-D-06). Taiwan thus discovered, within thirty months of building its modern constitutional court, both of the genus's known failure modes — court-paralysis by quorum statute and court-starvation by confirmation blockade — and survived them, so far, only by the contested expedient of the court adjudicating the rules of its own existence.

6. The 2024–2026 Constitutional Stress Test

6.1 The Legislative-Reform Package and the Bluebird Spring (May–October 2024)

The hung Eleventh Legislative Yuan's first major act was the legislative-reform package (國會改革法案, Guohui Gaige) of May 2024: amendments to the Law Governing the Legislative Yuan's Power and the Criminal Code that created a presidential state-of-the-nation address with legislator questioning; required officials to answer legislative questions without "counter-questioning," on pain of sanction; created investigative committees with document- and testimony-compulsion powers over officials, military, juristic persons, and private parties; and created a criminal offence of contempt of the legislature (藐視國會罪) covering official misrepresentation before the chamber. The bloc's framing was Sunflower-descended oversight finally institutionalised; the DPP's framing was an unconstitutionally acquired inquisition power passed without clause-by-clause deliberation. The procedural manner of passage — show-of-hands votes amid floor brawls, with text finalised in caucus rather than committee — generated the Bluebird movement (青鳥行動, named for the Qingdao East Road of the protest site): crowds outside the LY peaking around 100,000 on the 24 May 2024 vote night [TBD-VERIFY: crowd estimates], the largest civic mobilisation since the Sunflower occupation and explicitly genealogically linked to it.

President Lai promulgated the package while the EY returned it (the LY re-passed it); the DPP caucus, the EY, the president, and the Control Yuan all petitioned the Constitutional Court. In Judgment No. 9 of 2024 (113年憲判字第9號, 25 October 2024), the Court struck the package's core: the contempt-of-legislature criminal offence, the compulsion of the president's question-and-answer appearance, the sanction-backed answer duties, and most of the investigative-committee compulsion architecture were held unconstitutional in whole or part, with the LY's investigative power confirmed in principle but confined to chamber-resolution-based, judicially reviewable, proportionate exercise [TBD-VERIFY: the judgment's provision-by-provision disposition]. The ruling resolved the package and detonated the next phase: the bloc's response was not legislative redrafting but the December 2024 strike at the Court itself (Section 5.3).

6.2 The Budget Knife and the Recall Wave (2025)

The January 2025 FY2025 budget round converted the confrontation into fiscal attrition: aggregate cuts and freezes of approximately NT$207.5 billion [TBD-VERIFY: exact figure, per TW-D-06] falling with announced intent on the agencies of the DPP state — the Constitutional Court's and Control Yuan's operating budgets, the MAC, public media, and defence lines whose reduction drew explicit US commentary (TW-D-06). The Executive Yuan's counter-instruments — reconsideration requests, constitutional referral, and public framing of the cuts as governance sabotage — define the budget choreography that repeated, at lower amplitude, in the FY2026 cycle (TW-D-06).

The civic counter-move was the Great Recall (大罷免) of 2025: mass petition campaigns, civil-society-initiated and belatedly DPP-supported, against KMT district legislators, exploiting the 2016 liberalisation of the Public Officials Election and Recall Act (which had lowered recall thresholds to a 25-per-cent-turnout, simple-majority rule). The vote days — 26 July 2025 (24 KMT legislators and the suspended Hsinchu mayor) and 23 August 2025 (7 more) [TBD-VERIFY: exact dates and counts, per TW-D-06] — failed in every single district, leaving the bloc's majority intact and chastening the recall instrument's partisans. Institutionally read, the wave demonstrated three things at once: that the recall, designed as an accountability device for individual malfeasance, had been repurposed as a chamber-control device; that Taiwanese electorates will punish that repurposing; and that the constitution's only mechanism for re-deciding a deadlocked legislature mid-term — since dissolution requires a no-confidence vote nobody will move — is this blunt, district-by-district instrument that failed. The deadlock therefore runs to the January 2028 elections on its own steam, with the November 2026 local elections (TW-D-05) as the only intervening verdict.

6.3 The Referendum Instrument's Parallel Volatility

The referendum, the other direct-democracy valve, has traced the same arc from cage to weapon to anticlimax. The 2003 Referendum Act was born caged (the "birdcage law," 鳥籠公投, with prohibitive thresholds and a gatekeeping committee); the 2017 amendment flung it open (proposal and passage thresholds slashed, the threshold becoming 25 per cent of the electorate voting yes plus a majority); the November 2018 cycle saw ten questions ride the local elections, several explicitly counter-governmental (anti-same-sex-marriage, anti-DPP energy policy), and the 2019 re-amendment decoupled referendums onto their own biennial August date after the governing party concluded that coupled referendums were ungovernable. The December 2021 cycle (four questions, all failing) and the 23 August 2025 nuclear-restart question — which produced a large yes plurality that nonetheless failed the 25-per-cent threshold [TBD-VERIFY: the 2025 referendum question, vote totals, and threshold outcome] — confirmed the post-2019 pattern: decoupled referendums fail for turnout, so the instrument now functions mainly as a mobilisation ritual and agenda signal. The 2022 constitutional-amendment referendum's failure (Section 3.5) sits in the same file. Taiwan has thus equipped itself with every direct-democracy instrument in the catalogue — initiative, referendum, recall, and referendum-ratified amendment — and each has converged on the same equilibrium: usable for warfare, unusable for decision.

6.4 What the Stress Test Exposed

The 2024–2026 cycle is best read as a systematic audit of the 1997/2005 settlement's deferred questions, each returning with interest:

The court-composition vulnerability. Justice appointments require LY consent with no anti-deadlock device — no recess-appointment power, no caretaker mechanism, no automatic seating, no supermajority-with-fallback. A hostile majority can therefore shrink the bench indefinitely, and a quorum statute can weaponise the shrinkage. Judgment No. 2 of 2025 patched operability but not legitimacy: an eight-member bench, all appointed by DPP presidents, adjudicating DPP-vs-LY disputes is exactly the configuration the consent power existed to prevent, and the blockade is exactly the abuse the power enables. Both readings are held sincerely; the constitution adjudicates neither.
The strangulation asymmetry. A legislature that may cut but not add, facing an executive it cannot replace without facing dissolution, will discover the budget as its only safe siege engine — the Chen-era lesson industrialised in 2025–2026.
The unused valves. No-confidence (never passed), cohabitation (never attempted), dissolution (never triggered): the system's formal devices for resolving divided government have a thirty-year non-use record, which is itself an institutional fact — actors have priced them as mutually assured destruction, so conflict flows instead through courts, budgets, recalls, and streets.
The amendment lock under stress. Every fix for the above — confirmation anti-deadlock rules, budget-process redesign, a German-style constructive no-confidence, Yuan consolidation — requires amendment; amendment requires the iron cage to open; the cage will not open; therefore the system's pathologies are, for the foreseeable future, to be managed rather than repaired. And behind the procedural lock stands the strategic one: any amendment process, once open, becomes a sovereignty event (Section 7.3), which is precisely why many actors — Beijing and Washington included — quietly prefer the cage shut.

7. The Architecture in Comparative Perspective

7.1 The Last Five-Power Constitution

Taiwan operates the only Sun Yat-sen constitution in existence — the PRC having taken the Soviet path, and no other state having adopted the five-power design. Its defenders, from the orthodox KMT constitutional tradition through contemporary scholars of "mixed constitutionalism," make a case worth stating fairly: the design anticipated by decades the late-twentieth-century global proliferation of "fourth-branch" institutions — independent electoral commissions, audit offices, ombudsmen, civil-service commissions, human-rights commissions — that constitutional theory now treats as standard equipment for new democracies; Sun simply constitutionalised at the founding what others bolted on later, and he did so from an indigenous institutional vocabulary (the examination state, the censorate) rather than by Western transplant. The redundancy critique answers: the fourth-branch analogy flatters two bodies that lack the independence-generating appointment structures the analogy assumes (both are presidentially nominated and legislatively confirmed, hence hostage to exactly the partisan dynamics of 2005 and 2024); the Control Yuan's powers triplicate the LY's oversight, the prosecutors', and the discipline courts'; and the costs are not merely fiscal but systemic, since five branches mean more confirmation chokepoints, more inter-organ boundary disputes for the Constitutional Court, and more institutional surface area for divided-government warfare. The 2024–2026 record arms both sides: the Control Yuan petitioned against the legislative-reform package (vindicating, say defenders, its supervisory function) and was budget-strangled for it (demonstrating, say critics, its dependence).

7.2 Taiwan in the Semi-Presidential Family

Within comparative semi-presidentialism Taiwan is a canonical case — Elgie's datasets, the Shugart-Carey framework, and the dedicated Taiwan literature (Yu-Shan Wu's "semi-presidentialism under stress," Jung-Hsiang Tsai, Yeh Jiunn-rong's The Constitution of Taiwan) converge on a few placements. Against France: Taiwan took the Fifth Republic's president-premier duality but omitted its two stabilisers — the free dissolution power that lets a French president call the question, and the cohabitation convention that lets a French parliament answer it; Taiwan's conditional dissolution and untested cohabitation leave deadlock without exits. Against South Korea: Korea's single five-year presidential term and pure-presidential cabinet produce the "imperial presidency then lame duck" cycle, where Taiwan's re-electable four-year presidency with a fusible premier distributes blame downward and prolongs presidential authority — at the premier office's expense. Against the family's failure cases (Weimar, interwar and post-Soviet president-parliamentarism), Taiwan is the standing counterexample: a president-parliamentary system — the subtype the literature flags as democracy-endangering — that has consolidated, alternated power three times, and absorbed two divided-government sieges without regime crisis. Whether that record shows the subtype's dangers overstated, or shows Taiwan's exceptional civic and judicial buffers doing work the design does not, is a live scholarly question on which the 2024–2026 evidence has been claimed by both sides.

7.3 The Constitution as Identity Question

The deepest comparative singularity is not structural but existential: the ROC constitution is the only democratic constitution whose formal identity is claimed by, and legally entangled with, a state that aims to absorb its territory. The entanglement runs in all directions at once. For the independence tradition, the constitution is the coloniser's document — drafted in Nanjing while Taiwan bled in 1947, imposed under martial law, never ratified by the Taiwanese — and its replacement by a "new constitution" born in Taiwan is the unfinished business of democratisation; yet that tradition's own presidents (Chen after his 2003–2006 new-constitution and "second republic" gambits collapsed under American and domestic resistance; Lai, the once self-described "pragmatic worker for Taiwan independence") have governed by invoking it, because the formula "the Republic of China (Taiwan) is already a sovereign state under its constitution" is the only sovereignty claim that does not trigger the crisis thresholds. For the KMT tradition, the constitution is the republic itself — the unbroken legal personhood of 1912 — and its "one China" frame (the unamended territory clause, the Additional Articles' "prior to national unification" preamble, the one-country-two-areas statutory architecture) is both conviction and cross-Strait insurance. For Beijing, in a standing irony, the same residue is load-bearing: the PRC's "one China" legal position is easier to maintain while Taiwan's own constitution nominally agrees, which is why the 2005 Anti-Secession Law's triggers are read to include formal constitutional moves toward independence, and why PRC commentary polices Taiwanese constitutional debate more anxiously than Taiwanese statute-making. Tsai Ing-wen's and Lai Ching-te's formulation that the ROC and the PRC "are not subordinate to each other" (互不隸屬) — denounced by Beijing as "new two-state theory" — is constitutional politics conducted entirely at the level of interpretation, because the textual level is sealed (TW-F-02, TW-K-01, TW-K-03).

Hence the double reading of the amendment deadlock with which this document's Key Takeaways opened, and which comparative perspective sharpens rather than resolves. Stability: rigid constitutions are the norm among consolidated democracies (the US, Japan — no amendment since 1947 — and Germany's entrenched core all live with hard locks), and Taiwan's lock performs a unique additional service, taking the most dangerous question in Asia formally off the table; the 2005 framers, on this reading, built better than they knew. Ossification: Japan's unamended constitution at least matches its territory and its institutions work; Taiwan's misdescribes its country, embeds two branches nobody would re-found, and cannot patch demonstrated failure modes in its court and budget machinery — and a democracy whose fundamental law can be changed by no democratic process is storing, not avoiding, legitimacy risk. Both at full strength; neither falsifiable yet; the constitution, meanwhile, continues to be changed the only ways it can be — by interpretation, by statute, by practice, and by the accumulating fact of who its people have become.

8. Conclusion

The Republic of China constitution is the strangest success in modern constitutional history. Written for a continental state that lost its continent within two years; suspended by its own amendment procedure within five months of taking effect; preserved for four decades as the legal façade of a party-state that violated it daily; and then — against every comparative expectation — defrosted rather than discarded, rebuilt through seven amendment rounds into the vessel of one of Asia's most vigorous democracies, without a founding rupture, a constituent assembly, or a single day's interruption of legal continuity. The method's genius was that it made democratisation negotiable: the KMT could concede everything important while conceding nothing formal, and Taiwan could acquire self-government without declaring the independence that would have invited war. The method's price is the present: a five-branch architecture nobody would design today, a semi-presidential settlement with its hardest questions deferred, two minor Yuans suspended between irrelevance and abolition, and — since 2005 — an amendment rule that has converted all of these from reform agenda into permanent condition.

The 2024–2026 stress test, still unresolved at this document's cut-off, is the system examining itself under load. The honest interim findings: the formal machinery's conflict-resolution devices (no-confidence, dissolution, cohabitation) remain unused because every actor prices them as catastrophic; conflict therefore flows through the Constitutional Court, the budget, the recall, and the street, each of which has been damaged by the traffic; and the system has nonetheless held — courts ruling, budgets enacted, recalls counted, soldiers in barracks — which is more than the comparative record of president-parliamentarism under divided government entitled anyone to expect. Whether that holding reflects deep consolidation or accumulated luck, and whether a constitution that can no longer be amended can indefinitely govern a country it no longer describes, are the questions this corpus will revisit as the 2026 local and 2028 national cycles answer them. What can be said with confidence is only this: Taiwan's constitutional order has survived every previous prediction of its impossibility — drafted for the wrong country, suspended at birth, democratised by its suspenders' heirs, and locked shut by its own democrats — and its career to date counsels humility about the next prediction.

Sources

Constitution of the Republic of China (1947), original text — including the preamble; Chapters I–XIV; Articles 4 (territory), 53 (Executive Yuan as highest administrative organ), 55–57 (premier appointment and EY-LY responsibility, original design), 62–76 (Legislative Yuan), 77–82 (Judicial Yuan), 83–89 (Examination Yuan), 90–106 (Control Yuan), 174 (amendment) — and the Additional Articles of the Constitution of the Republic of China (中華民國憲法增修條文), consolidated text as amended through 10 June 2005, including the preamble's "prior to national unification" formula.
Temporary Provisions Effective During the Period of National Mobilisation for Suppression of the Communist Rebellion (動員戡亂時期臨時條款), adopted 10 May 1948, as revised 1960, 1966 (twice), and 1972; abolished with the termination of the Period of National Mobilisation declared by President Lee Teng-hui effective 1 May 1991.
Judicial Yuan, Council of Grand Justices / Constitutional Court. Interpretation No. 31 (1954, continuation of the first-term parliament); No. 76 (1957, the tri-organ parliament doctrine); No. 261 (21 June 1990, retirement of the senior parliamentarians by 31 December 1991); No. 499 (24 March 2000, voiding the Fifth-round amendments); No. 520 (15 January 2001, the Fourth Nuclear Power Plant suspension); No. 632 (2007, the Control Yuan confirmation deadlock); No. 748 (24 May 2017, same-sex marriage); Constitutional Judgment No. 9 of 2024 (113年憲判字第9號, 25 October 2024, the legislative-reform package); and Constitutional Judgment No. 2 of 2025 (114年憲判字第2號, 28 March 2025, the Constitutional Court Procedure Act amendment) [TBD-VERIFY: 2025 case number and date].
Constitutional Court Procedure Act (憲法訴訟法), in force 4 January 2022, and the December 2024 amendment thereto raising participation and decision thresholds; Law Governing the Legislative Yuan's Power (立法院職權行使法) as amended May 2024; Criminal Code amendment creating the contempt-of-legislature offence (May 2024); Referendum Act (公民投票法) 2003, as amended December 2017 and June 2019; Public Officials Election and Recall Act recall provisions as amended 2016; Act Governing Relations Between the People of the Taiwan Area and the Mainland Area (1992); Organic Acts of the Control Yuan and Examination Yuan as amended 2020 (including the National Human Rights Commission Organic Act 2020).
Sun Yat-sen. The Three Principles of the People (三民主義, 1924 lectures) and Fundamentals of National Reconstruction (建國大綱, 1924); the May Fifth Draft Constitution (五五憲草, 1936); records of the 1946 Political Consultative Conference and the Constituent National Assembly, Nanjing, November–December 1946.
Chang, Carsun (Zhang Junmai). The Third Force in China (Bookman Associates, 1952), and biographical-constitutional studies of Chang's drafting role, including Roger B. Jeans, Democracy and Socialism in Republican China: The Politics of Zhang Junmai (Rowman & Littlefield, 1997).
Yeh, Jiunn-rong. The Constitution of Taiwan: A Contextual Analysis (Hart Publishing, 2016) — the standard English-language treatment of the constitutional order, the amendment rounds, and the Additional Articles method.
Wu, Yu-Shan. "Semi-Presidentialism — Easy to Choose, Difficult to Operate: The Case of Taiwan," in Robert Elgie and Sophia Moestrup (eds.), Semi-Presidentialism Outside Europe (Routledge, 2007); Shugart, Matthew Søberg and John M. Carey, Presidents and Assemblies: Constitutional Design and Electoral Dynamics (Cambridge University Press, 1992); Elgie, Robert, Semi-Presidentialism: Sub-Types and Democratic Performance (Oxford University Press, 2011); Tsai, Jung-Hsiang, "Sub-types of Semi-presidentialism and Political Deadlock," French Politics 6 (2008).
Roy, Denny. Taiwan: A Political History (Cornell University Press, 2003) — the Temporary Provisions era, the ten-thousand-year parliament, and the Lee-era reforms; Rigger, Shelley, Politics in Taiwan: Voting for Democracy (Routledge, 1999) and Why Taiwan Matters (Rowman & Littlefield, 2011).
Cheng, Tun-jen and Stephan Haggard (eds.). Political Change in Taiwan (Lynne Rienner, 1992); Tien, Hung-mao, The Great Transition: Political and Social Change in the Republic of China (Hoover Institution Press, 1989) — the late-authoritarian constitutional structure and the National Affairs Conference.
Lin, Jih-wen; Batto, Nathan F.; and the electoral-reform literature on the 2005 round: Batto, Nathan F. et al. (eds.), Mixed-Member Electoral Systems in Constitutional Context: Taiwan, Japan, and Beyond (University of Michigan Press, 2016); Huang, Chi, "Electoral System Change and Its Effects on the Party System in Taiwan," in the same volume.
Yeh, Jiunn-rong and Wen-Chen Chang. "The Emergence of East Asian Constitutionalism: Features in Comparison," American Journal of Comparative Law 59 (2011); Chang, Wen-Chen, "Courts and Judicial Reform in Taiwan: Gradual Transformations Towards the Guardian of Constitutionalism and Rule of Law," in Jiunn-rong Yeh and Wen-Chen Chang (eds.), Asian Courts in Context (Cambridge University Press, 2014).
Central Election Commission (中央選舉委員會). Official returns: presidential elections 1996–2024; legislative elections 1992–2024; the 26 November 2022 constitutional-amendment referendum on the voting age; the 2018, 2021, and 2025 referendum cycles; and the 26 July and 23 August 2025 recall votes [TBD-VERIFY: 2025 figures].
Office of the President (ROC) and Executive Yuan. Documentary record of the National Affairs Conference (1990), the National Development Conference (1996), the termination declaration of 30 April–1 May 1991, and the presidential and EY statements across the 2024–2026 inter-branch confrontation; Legislative Yuan, Legislative Yuan Gazette (立法院公報), procedural records of the May 2024 reform-package passage, the December 2024 justice-confirmation votes and Constitutional Court Procedure Act amendment, and the FY2025–FY2026 budget proceedings.
Control Yuan (監察院). Annual reports, impeachment and corrective-measure records, and the National Human Rights Commission's reports 2020–2026; Examination Yuan (考試院), organic-reform documentation 2020 [TBD-VERIFY: post-2020 membership and term figures].
Templeman, Kharis. Hoover Institution Taiwan programme commentary on the 2024–2026 divided-government and Constitutional Court confrontation; Chen, Yu-Jie (Academia Sinica), commentary on the Constitutional Court Procedure Act amendment and confirmation blockade; Nachman, Lev and Sung, Wen-Ti, contemporaneous analysis of the Bluebird movement and the 2025 recall wave (cross-referenced in TW-D-06).
Kerr, George H. Formosa Betrayed (Houghton Mifflin, 1965) and the 228-era documentary record (per TW-J-01) for the constitutional founding's Taiwan context; Phillips, Steven E., Between Assimilation and Independence: The Taiwanese Encounter with Nationalist China, 1945–1950 (Stanford University Press, 2003).
Bush, Richard C. Untying the Knot: Making Peace in the Taiwan Strait (Brookings, 2005) and Difficult Choices: Taiwan's Quest for Security and the Good Life (Brookings, 2021) — the cross-Strait stakes of constitutional change, the Chen-era new-constitution episode, and US policy responses; the PRC Anti-Secession Law (10 March 2005).
Freedom House, Freedom in the World Taiwan country reports 2020–2026; International IDEA and Verfassungsblog / I·CONnect comparative commentary on the December 2024 Constitutional Court Procedure Act amendment and the court-paralysis question, 2024–2026.
Taipei Times, United Daily News (聯合報), Liberty Times (自由時報), China Times (中國時報), Central News Agency (CNA), Focus Taiwan, Reuters Taipei bureau, and The Economist. Print and digital archives, 1991–June 2026, on the seven amendment rounds, the provincial freezing, the 2005 National Assembly, the 2022 referendum, the Bluebird movement, the justice-confirmation blockade, and the 2025 recall and referendum cycles.

TW-A-02: Cross-Strait Relations (1949–2024) — the cross-Strait frame within which every constitutional choice from the Additional Articles method to the amendment freeze has been made.
TW-A-03: Chen Shui-bian Presidency (2000–2008) — the first divided-government cycle, the revolving-door premiership, and the new-constitution episode treated institutionally in Sections 4.2 and 7.3.
TW-B-01: 1987 Lifting of Martial Law and Democratisation — the dismantling of the parallel state whose constitutional layer this document treats in Section 2.3.
TW-C-01: Ma Ying-jeou Presidency (2008–2016) — the unified-government configuration of Section 4.3, including the 2013 Ma-Wang strife.
TW-D-02: Lai Ching-te Presidency (2024–2025) — the chronological narrative of the divided-government confrontation whose institutional anatomy is Sections 5.3 and 6.
TW-D-06: Taiwan 2026 Budget Battle — KMT-TPP Bloc Confrontation and Defence Spending — the FY2025–FY2026 budget cycles, the justice-confirmation paralysis continuation, and the post-recall equilibrium on which this document's Section 6 relies.
TW-J-01: Three Accounts — The 228 Incident (1947) and Its Memory Politics — the founding violence contemporaneous with the constitution's promulgation, and the memory politics that shape the "coloniser's document" reading in Section 7.3.
TW-K-01: Tsai's 2019 "Five Musts" Rejection of the Xi Address — the interpretive-level constitutional politics of the "not subordinate to each other" line.
TW-K-03: The 7 November 2015 Ma-Xi Singapore Meeting Decision — the "one China, respective interpretations" diplomacy conducted on the constitutional residue Section 7.3 describes.
TW-N-01: Taiwan in International Perceptions (1971–2026) — the external readings of Taiwan's constitutional-identity question and democratic credibility.
TW-G-05: back-reference added by symmetry sweep
TW-H-PRES-01: back-reference added by symmetry sweep
TW-J-02: back-reference added by symmetry sweep
TW-M-01: back-reference added by symmetry sweep
TW-O-01: back-reference added by symmetry sweep